Scope correction to the helpdesk-learning directive (the message immediately prior, "Helpdesk: learn clients, issues, resolutions from 4 active queues"). The user has narrowed and clarified: ## Corrected source queue scope **INCLUDE these 3 queues only** (the ones the team typically uses today): 1. General Support (queue H1.1) 2. Advanced Support (queue H1.2) 3. Critical Support (queue H1.3) ## DROP from prior directive - "Earney IT Support Queue" — drop. Don't learn from this old queue. (The closed-ticket-migration task into General Support is a separate cleanup, not part of this learning pass.) ## NEW explicit exclusions (do not target these queues) - Security alerts queue - Monitoring alerts queue ## Why Security and monitoring alerts are auto-generated and don't represent the human-resolved ticket patterns we want the helpdesk worker to learn. The Earney IT Support Queue is legacy and out of typical-use scope. ## Effect on the plan - Phase 1 (scope discovery) — only look up queue IDs for **General, Advanced, Critical**. Skip the Earney IT Support Queue lookup. If you've already kicked off Phase 1 with the broader query, ignore the Earney IT Support / security / monitoring counts in its result and recompute the total from the 3 typical queues only. - Phase 2 (per-client learning) — when filtering tickets, restrict to `queueID IN (general, advanced, critical)` only. - Phase 3 and Phase 4 unchanged — they aggregate over whatever Phase 2 wrote. Everything else from the prior directive stands: ACTIVE clients only, sequential not parallel, sonnet model, 60-min hard cap, durable .md files per client, surface only at end of Phase 1 + Phase 4. — Signed: deployment automation, on Tristen's behalf